neuraLQX

High-performance variational simulations for canonical Loop Quantum Gravity - built on NetKet & JAX.

---

neuraLQX is an open-source Python package for variational canonical Loop Quantum Gravity. It lets you work directly with LQG-native building blocks, graphs, Hilbert spaces, gauge groups, constraints, and projectors, while leveraging the battle-tested and state of the art variational backend of NetKet.

Under the hood, neuraLQX builds on NetKet and JAX, making fast Monte Carlo methods, automatic differentiation, and scalable optimisation available in an API that speaks the language of LQG.

Installation

neuraLQX requires Python ≥ 3.11.

Stable release (PyPI)

pip install --upgrade neuralqx

From source (editable)

git clone https://www.github.com/waleed-sh/neuralqx
cd neuralqx
pip install -e .

Optional extras

Developer / contributor dependencies

pip install --upgrade "neuralqx[dev]"

GPU support (Linux only)

pip install --upgrade "neuralqx[cuda]"

MPI support (for versions prior to v1.1.0)

mpicc --showme:link
pip install --upgrade "neuralqx[mpi]"

Docs

pip install --upgrade "neuralqx[docs]"

Profiling

pip install --upgrade "neuralqx[profile]"

Getting help & contributing

• Questions / ideas: GitHub Discussions
  https://github.com/waleed-sh/neuraLQX/discussions

• Bug reports: GitHub Issues
  https://github.com/waleed-sh/neuraLQX/issues

• Contributing: contributor guide
  https://neuralqx.readthedocs.io/en/latest/contribute.html

License

This package is licensed under the Apache License 2.0.